. 2
( 7)


Norway 29.9
Canada 23.0
New Zealand 21.5
United Kingdom 19.2
Ireland 18.8
United States 17.2
Australia 17.2
Japan 15.9
Italy 11.6
Germany 11.3
Netherlands 9.31
Finland 4.69
France 3.41
Austria 1.83
Belgium 0.57
Spain 0.11
Portugal 0.11
Greece 0.04
Sources: Expenditure data from OECD 2003a; GDP
and demographic data from OECD 2003b.

and Portugal spend the least on the elderly. Scandinavian welfare states,
contrary to their reputation as big spenders, actually spend only moderately
on the elderly on a per capita basis.
The key spending category for working-age adults, spending on active
and passive labor market programs per registered unemployed person
(Table 2.2), ranges from a high of well over one and a half times the GDP per
capita in Denmark and Sweden to a low of just over 20 percent of GDP per
capita in the United States and Greece. Standardizing aggregate unemploy-
ment expenditure ¬gures by the number of registered unemployed, rather
than by the number of unemployment insurance bene¬ciaries, allows for
an estimate of the extension as well as the level of unemployment bene¬ts.
Countries that have high numbers of uninsured unemployed people (e.g.,
¬rst-time job seekers) or that spend very little on unemployment bene¬ts
and active labor market policies will have low per capita spending, and vice
Measuring the Age of Welfare

Table 2.6 Per capita health spending ratios

Spending Source of Spending Source of
Country ratio health Country ratio health
(year) (65+/0“64) spending (year) (65+/0“64) spending
U.S. (1987) 8.9 Public only Finland 4.0 Public and
(1990) private
Japan (1997) 4.9 Public only Netherlands 3.9 Public and
(1994) private
U.S. (1995) 4.6 Public and U.K. (1997) 3.4 Public and
private private
Ireland (1979) 4.5 Public only France (1991) 3.0 Public and
Canada (2000) 4.5 Public only Sweden 2.8 Public and
(1990) private
New Zealand 4.3 Public only Germany 2.7 Public only
(1998) (1994)
Denmark 4.1 Public only Italy (1983) 2.2 Public only
Australia 4.0 Public and Portugal 1.7 Public and
(1989) private (1993) private
Sources: OECD 1998; 2003b; U.S. “public only” ¬gure calculated from Waldo, Sonnerfeld,
MacKusick, and Arnett 1989.

Labor market supports may of course be targeted at quite different age
groups across countries and across time. For example, in 1996 Germany
introduced an extension of unemployment bene¬ts for workers transition-
ing into retirement, while at the same time reducing the period of eligibility
for younger workers. OECD ¬gures on active labor market policy spending
include outlays for early retirement “for labor market reasons,” as well as
youth job training programs. A detailed survey of the bene¬ciaries of labor
market policies could be undertaken in order to “allocate” spending to cat-
egories of older and younger workers (chapter 5 undertakes such a study for
Italy and the Netherlands). For the present, it is enough to note that coun-
tries vary widely in their average generosity toward the unemployed, even
taking into account programs that do not ¬nd their way into comparisons
of the replacement rates of standard unemployment insurance bene¬ts.
The Scandinavian countries resume their accustomed position as welfare
leaders when we turn to labor market policy spending directed toward
working-age adults. But the Nordic countries are not alone in allocating
resources generously to the unemployed; the Netherlands and Belgium are
Age in the Welfare State

also among the top ¬ve in this regard. Ireland, New Zealand, and Canada
belie their reputation as residualist Anglo-Saxon welfare states, appearing
squarely in the middle of the pack on labor market spending per unemployed
person. Spain, Italy, and Greece, which spend lavishly on programs for the
elderly, have among the least generous welfare states for the unemployed.
The data for occupational health and safety expenditures (Table 2.3) are
spottier, but display a similar pattern.
Direct expenditures on bene¬ts for families with children (Table 2.4) “
child allowances and services such as child care, public summer camps, or
mothers™ aides “ follow a pattern similar to bene¬ts geared toward partici-
pants in the labor market. (The main exceptions are Canada and the Nether-
lands, which rank substantially lower on spending per child than they do on
labor market policy spending, and Australia, which ranks higher.) The over-
all similarity in the country rankings for spending on labor market supports
and on families is perhaps surprising. There is no inherent link between
the two types of policies other than the fact that they are both aimed at
non-elderly bene¬ciaries. Perhaps one could make the case that spending
on children is a form of labor market support if it allows women to enter
the work force, as in Sweden and Norway. Yet some of the welfare states
of Continental Europe not renowned for their efforts to support female
labor force participation nevertheless spend quite generously on families
with children. Spending on families with working-age heads, both with and
without children, seems to follow similar patterns, which quite evidently
differ from the pattern of spending on the elderly.
But not all direct public spending on families with children comes
through the paradigmatic social insurance institutions of the welfare state.
Education spending, which is often ignored in comparative studies of the
welfare state, undoubtedly “counts” as social spending, albeit of a differ-
ent kind, and is clearly focused on the non-elderly. Table 2.5 shows pub-
lic expenditures on primary, secondary, and tertiary education per school-
age person, as a percentage of GDP per capita, averaged over the period
1992“8.5 Perhaps even more than is the case for social insurance programs,
it is dif¬cult to know whether public spending on school construction,
teachers™ salaries, or high-end scienti¬c equipment is really a good mea-
sure of how much education is being provided to a nation™s children and
young adults. In the United States, for example, education spending is

5 These are the years for which cross-national and cross-time comparable data are available
from the OECD.

Measuring the Age of Welfare

concentrated at the tertiary level, re¬‚ecting a strong emphasis on research
technology rather than on teachers and classrooms for primary and sec-
ondary school students. Caution in interpreting these data is clearly
required. Still, the ¬gures for per capita education spending highlight the
general tendency for countries that are elderly-oriented in the ¬eld of social
insurance to spend relatively little on the non-elderly in the form of edu-
cation, and vice versa.
Health spending, like education spending, is a very large component
of social spending in OECD countries, ranging from 20 to 40 percent of
total social spending, or around 5 to 8 percent of GDP in 1998 (OECD
2003b).6 Health spending per capita on the elderly and non-elderly, like
social insurance and education spending, varies in important ways across
countries. While in some countries providing an adequate standard of care
to children and pregnant women is the most basic test of the health system,
elsewhere either health bene¬ts are not publicly provided at all to non-
poor, non-elderly citizens or access to speci¬c bene¬ts varies by age group.
In still other countries, where health care is a universal bene¬t, in practice
rationing may lead to unequal emphases on treatment for elderly and non-
elderly patients.
What is an appropriate age-sensitive measure of public health spending?
The OECD has collected from many of its member countries statistics on
health care spending by age group (see Table 2.6). It bears emphasizing,
however, that these statistics are incomplete, widely disparate in terms of
the years and populations covered, and in some cases include private as well
as public health spending. For those countries where the per capita spend-
ing ratio reported by the OECD includes both public and private health
spending, we can make the simplifying assumption that per capita health
spending ratios by age groups are the same in both the public and private
health sectors. Because private health expenditures are a very small portion
of the total in most countries, this assumption is in most cases unproblem-
atic. In the United States, where private health spending is important and
serves a younger population than do the publicly provided Medicare and
Medicaid programs, per capita spending ratios for both public only and
public and private health expenditures are shown.
While per capita health expenditures on elderly and non-elderly groups
continue to highlight the United States and Japan as among the most

6 Note that with the exception of this paragraph, I use the term “total” social expenditures
throughout this book to refer only to nonhealth expenditures.

Age in the Welfare State

elderly-biased countries, Italy appears dramatically more youth-oriented
in its health spending than in other areas of the welfare system. France
and Germany also seem to be more youth-oriented than one might have
guessed based on the social insurance expenditure data. The Scandinavian
countries, on the other hand, give more emphasis than expected to older
patients. Given the diversity of measurement techniques and sources used
in deriving the health spending ratios, and the problematic nature of health
spending as a measure of health care emphasis for different age groups
(Meyer and Moon 1988), how seriously should one take the per capita
health spending ratio as a measure of the age orientation of health policy?
Some support for the measure is provided by the observation that the
very low ¬gures observed for Italy and Portugal correspond to a known
property of Southern European welfare states: universalist health systems
counter the fragmentation and strati¬cation characteristic of other areas
of social provision in these systems (Ferrera 1996c; Gough 1996).7 One
need not rely solely on the per capita health spending ratio as a measure
of age orientation in health policy, though. Changes in infant mortality
rates can be used to evaluate whether the per capita health spending ratio
re¬‚ects real differences in the distribution of health resources to elderly and
non-elderly populations. Infant mortality is sensitive to levels of prenatal
care and health status of pregnant women. If changes in infant mortality
are inversely related to per capita health spending ratios, we can be more
con¬dent in the validity of the health spending ratio as a measure of age
orientation of health policy. In fact, when controlling for the wealth of a
nation, a lower per capita health spending ratio (i.e., more spending on the
non-elderly relative to the elderly) is strongly associated with declines in
infant mortality.8

7 On the other hand, recent changes to the Italian health system exempt elderly people,
regardless of income, from many co-payments. This suggests that even within an excep-
tionally age-neutral health subsystem there may be pressures toward conformity with the
overall elderly bias of social provision in Italy.
8 GDP per capita is used as an instrument for the level of infant mortality at the start of
the period under study. Results of the OLS regression with percentage change in infant
mortality 1985“98 as the dependent variable are as follows:
Coef¬cient Standard error
Constant 1.238
Health spending ratio 0.898 0.251
GDP per capita, 1985 0.110 0.099
(millions $US PPP)

Measuring the Age of Welfare

This result lends credence to the claim that per capita health spend-
ing ratios are a good measure of the emphasis of different welfare states
on health care for elderly and non-elderly populations. It further suggests
that case studies of the development of health programs may be useful in
illuminating the causal processes behind the development of particular age
orientations in social policy regimes: health policies in Southern European
countries may well be exceptions that prove the rule when compared with
other kinds of social programs.

Summarizing the Age Orientation of Direct Expenditures
So far we have considered the relative generosity of direct public expen-
ditures for a variety of social needs “ income supports and services for the
elderly, the unemployed, and families with children, as well as education
and health care “ in OECD countries. And we have detected some evi-
dence that across different policy areas, countries are consistent in the way
that they allocate resources to different age groups in their populations.
Clearly, some countries spend more on services and others more on cash
bene¬ts; some more on active labor market policies and others more on
unemployment bene¬ts; some try to ensure equality of access via educa-
tion, some through the labor market; and others equality of outcome via
income supports. Among those countries that spend a lot on their non-
elderly populations, some allocate more resources to children and others
to working-age adults. Leaving aside these ¬ner distinctions, however, a
global measure of the age orientation of all direct social spending can give
us valuable information about policy priorities with regard to different age
groups in different countries. Such a measure based on direct expenditures,
if it is consistent with what we know about how other policy instruments
treat different age groups, also offers a highly tractable way to analyze the
age orientation of welfare states.
Let us construct a measure that is a summary comparison of public social
insurance expenditures on the elderly (aged 65+ or in formal retirement)
and expenditures on the non-elderly (children and adults aged 0“64 and not
in formal retirement). We can call this measure the Elderly/Non-elderly
Spending Ratio, or ENSR. Several types of direct expenditures must unfor-
tunately be excluded from this measure. The largest is cash bene¬ts for the
disabled, which accounts for between 2 and 11 percent of direct social
expenditures excluding health care in OECD countries (OECD 2003b).

Age in the Welfare State

Table 2.7 Elderly/non-elderly spending ratio
(ENSR) (average 1985“2000)

Denmark 5.75
Sweden 6.50
Ireland 7.11
Belgium 8.32
Finland 8.86
Australia 9.29
Norway 9.89
Netherlands 10.2
United Kingdom 10.4
New Zealand 11.4
France 12.9
Canada 14.0
Spain 15.7
Germany 16.0
Austria 17.4
Portugal 18.6
Greece 24.7
Italy 28.9
United States 38.5
Japan 42.3
Sources: Spending data from OECD 2004; demo-
graphic data from OECD 2003b.

Disability pensions are converted into retirement pensions upon reaching
the retirement age in some countries, which means that virtually 100 per-
cent of this category could be safely designated as non-elderly spending.
But in other countries disability pensions can be cumulated with retire-
ment pensions, complicating our accounting. Similarly, in most countries
it is not possible to track the age of bene¬ciaries of social transfers for hous-
ing. Cash bene¬ts are probably one of the least important determinants of
housing outcomes, though, so we consider housing policy separately from
direct social expenditures. Finally, a paucity of reliable data on occupational
injury and sickness spending as well as public health spending leads us to
exclude these categories from the ENSR. Given the dif¬culties in inter-
preting aggregate public education spending, it seems wise to compare a
measure that does not include it with one that does.
Table 2.7 shows values of the ENSR for twenty OECD countries,
averaged over the years 1985 through 1998. Non-elderly expenditures
include payments for unemployment bene¬ts (including early retirement
Measuring the Age of Welfare

for labor market reasons), active labor market policies, family cash bene¬ts,
and services for families. Elderly expenditures are those for old-age pen-
sions (including early retirement for non“labor market reasons), survivors™
pensions, and services for the elderly and disabled (primarily nursing
homes). Spending on the elderly is divided by the number in the popu-
lation aged 65 or more; spending for families with children and spending
on active and passive labor market policies are divided by the population
aged less than 65.
The numerical values for the ENSR represent, in a strict sense, a spend-
ing ratio. But it is inadvisable to conclude from the ENSR that, for example,
the United States spent almost forty times as much on the elderly as on the
non-elderly on a per capita basis, while Denmark spent only ¬ve times as
much. In the ¬rst place, this basic measure is not a complete survey of expen-
ditures in all areas. Second, while almost all elderly people living in OECD
countries have access to some form of public pension, not all persons under
age 65 are, at any given time, receiving bene¬ts such as child allowances
or unemployment insurance.9 Furthermore, the value to the individuals
concerned of different types of payments may not be captured by the mag-
nitude of spending. Supports for children and working-age adults typically
aim to supplement market incomes or replace them for a short period of
time. All other things being equal, we would not expect them to be as costly
on aggregate as old-age pensions, which provide a year-round replacement
for market income.
Incorporating per capita education spending ¬gures into the basic
ENSR, as in Table 2.8, reduces both the levels and the variation across
countries considerably from the basic ENSR. Education spending is large
enough to buffer the effects of other social spending even for the countries
that spend least on their schools. The major shift in ENSR rankings after
including education spending occurs in Japan, which seems to spend heavily
enough on education to modify its strong elderly orientation in the ¬eld of
social insurance.
Raw per capita spending on different categories of bene¬ciaries such
as the elderly or the unemployed, the ratio of health spending on differ-
ent age groups, the basic ENSR, and the ENSR with per capita education

9 An alternative measure using the number of children and the unemployed as the divisor
for non-elderly expenditures can be constructed, but it is more dif¬cult to interpret and
does not yield signi¬cantly different results. Only Austria ranks signi¬cantly differently (less
elderly-oriented) using this measure.

Age in the Welfare State

Table 2.8 ENSR with education spending
(average 1985“2000)

Denmark 0.32
Ireland 0.40
Netherlands 0.40
Sweden 0.44
Norway 0.44
Australia 0.53
New Zealand 0.56
Belgium 0.65
Canada 0.70
Finland 0.72
Germany 0.75
United Kingdom 0.76
Japan 0.83
Portugal 0.87
Austria 0.97
France 1.07
United States 1.24
Italy 1.66
Spain 1.72
Greece 2.53
Sources: Expenditure data from OECD 2003a; 2004;
demographic data from OECD 2003b.

expenditures all present slightly different views of the variety of social policy
orientations across OECD countries, through the lens of direct expendi-
tures. Those countries clustered at the middle of the spectrum on the basic
ENSR measure seem to employ different combinations of policies in order
to attain a generally age-balanced policy pro¬le. However, with the possi-
ble exception of health care, these measures of different policy expenditure
combinations generally point toward the same conclusions: Italy, Greece,
the United States, and Japan are heavily elderly-oriented countries, with
Spain, Portugal, and Austria not far behind in most respects. On the other
hand, the Netherlands, the Scandinavian countries, and some of the Anglo-
Saxon world (the United Kingdom, Ireland, Canada, and New Zealand)
provide a more balanced repertoire of direct social services and bene¬ts to
different age groups in the population.
It is worth emphasizing once again that the age orientation of social
spending does not seem to correspond with typologies of welfare states
based either on welfare state “effort” (the level of aggregate social spending)
Measuring the Age of Welfare

or welfare state “worlds.” High-spending countries may be either elderly-
oriented (Italy) or quite youth-oriented (the United Kingdom); low-
spending countries may also be elderly-oriented (the United States) or
youth-oriented (Australia). Most strikingly, many of Esping-Andersen™s
Liberal countries, which score low on his decommodi¬cation index because
of the presence of means-tested bene¬ts, are as youth-oriented as the truly
universalist countries of Scandinavia. High spending and even the decom-
modifying power of the welfare state do not then divide the youth-oriented
from the elderly-oriented welfare states. Rather, the age-orientation of the
welfare state depends on the target of social spending, on who gets decom-

Re¬ning the Measure: Taxes and the “Hidden Welfare State”
Direct expenditures on welfare goods tell one story about the extent to
which different age groups bene¬t from current social programs. But social
policy is made up of more than direct welfare expenditures (Howard 1997).
Tax systems both effectively reduce social spending, through taxes on cash
bene¬ts, and increase it, through tax expenditures on major social programs.
The distributive effects of speci¬c tax policies are notoriously dif¬cult to
interpret, which is of course one reason why they play such a prominent
role in social policy. Good household-level tax and transfer data would be
the most ef¬cient and probably the most accurate way of determining the
comprehensive tax-bene¬t position of different kinds of families. However,
even the most rigorously standardized comparative micro-data sets10 con-
tain limited (and not terribly reliable) information on taxes paid by house-
holds. Until better household-level data become available, aggregate tax
expenditure data provide the best estimates available of the age orientation
of tax policies.
Very signi¬cant perils confront those analysts who would compare tax
expenditure data across countries and across policy areas within a given
country (Adema, Pearson, Einerhard, et al. 1997). Adema et al. provide
some of the only truly comparable data on tax expenditures on social welfare
programs in OECD countries, reported here in Table 2.9. Of the six nations
included in their study, only the United States and the United Kingdom

10 The Luxembourg Income Study project and European Community Household Panel are
two sources of pre- and post-tax household income for OECD countries.

Age in the Welfare State

Table 2.9 Social-¬scal measures as a percentage of GDP, 1993

Direct taxes Social-¬scal
Total and social measures on
social-¬scal Direct social contributions old-age
measures expenditures paid on transfer bene¬ts
Country (% of GDP) (% of GDP) income (% of GDP)
Denmark 0.08 30.5 3.91 0.08
Germany 0.86 28.7 2.57 0.08
Netherlands 0.76 30.6 5.86 0.68
Sweden 0.20 38.3 5.31 0.20
U.K. 3.03 23.4 0.19 2.68
U.S. (federal 2.00 15.0 0.08 0.85
level only)
Source: Adema et al. 1997.

have tax expenditures on social policy that are signi¬cant compared with
the magnitude of direct social expenditures. This is true even taking into
account the effective reduction in direct expenditures due to direct taxation
of social bene¬ts, which can amount to around 3 to 6 percent in Northern
European countries. For example, even after tax clawbacks on income such
as pensions and unemployment insurance, Germany still spends roughly
26 percent of GDP on direct spending for social programs, as compared
with less than 1 percent on indirect spending (tax expenditures).
Happily, the countries where tax expenditures on social policy are sig-
ni¬cant compared with direct expenditures report quite comprehensively
on tax expenditures. As a result, it is possible to con¬rm whether the social
policy delivered through taxation mechanisms in these countries has the
same general age orientation as policy carried out through direct expendi-
tures. Despite the risks inherent in comparing tax expenditure data across
different policy areas and different countries, it is worth examining the tax
expenditure ¬gures for those countries where they may be expected to play a
large part in social policy: the United States, the United Kingdom, Canada,
and Australia. How much goes to the elderly in the form of tax expenditures
on private pensions or special health insurance programs?11 How much do

11 Classifying all tax expenditures on private pensions as an elderly expenditure is admittedly
somewhat arbitrary, since the age of the average bene¬ciary will depend on whether the
tax relief is granted at the time of the payment into the pension plan or at the time of
liquidation of the pension. Since most countries do both, it is very dif¬cult to judge which
is the most reasonable assumption on balance. For the sake of consistency, I choose to

Measuring the Age of Welfare

Table 2.10 Tax expenditures on the elderly and non-elderly (billions of national currency)

Tax expenditures on:
Unemployment, allowances, ENSR for tax
Country Old age labor market Education child care expenditures
U.S. (1995) 89,885 7,245 2,785 8,735 4.79
U.K. (1993“4) 18,120 3,500 550 1,450 3.29
Canada (1992) 17,390 4,471 954 2,945 2.08
Australia (1993“4) 46,423 1,530 21 182 26.8
Source: OECD 1996.

families and young people gain from tax exemptions on unemployment
bene¬ts, child care, or family allowances? Table 2.10 allows for some very
tentative judgments along these lines.
Adema et al.™s data on tax expenditures in the United States, United
Kingdom, Canada, and Australia provide an interesting complement to the
direct-expenditure measures of the age orientation of social policy in these
Liberal welfare states. Australia saw a sharp decrease in the share of direct
public expenditures devoted to the elderly following privatization of its
public pension system in 1986. The data in Table 2.10, which reveal a very
large imbalance in favor of the elderly in indirect social bene¬ts, indicate
that the overall balance between age groups may not have changed all that
much in Australia since 1986. Rather, subsidization of the elderly seems to
be increasingly undertaken through the tax system, while support for the
non-elderly continues to ¬‚ow through direct expenditures. A time series in
tax expenditures dating from before the pension reform would be invaluable
in con¬rming or refuting this possibility.
If the relative youth orientation of the Australian system may be
explained away by the continued presence of policy aids for the elderly
in the form of tax bene¬ts, the opposite seems to be true of Britain. Britain
in 1993 reported tax expenditures on social policy on the order of 3 per-
cent of GDP, while its direct social expenditures were around 23 percent of
GDP. So the relatively youth-oriented social policy orientation indicated by
the ENSR for direct expenditures in the United Kingdom is not canceled

classify tax expenditures on private pensions as elderly-oriented, but it is well to keep in
mind that these tax expenditures may also have an immediate impact on the disposable
income of non-elderly persons.

Age in the Welfare State

out, as in Australia, by large tax expenditures that heavily favor the elderly.
While it is true that subsidies for private pensions make up the lion™s share
of tax expenditures in Britain, the disparity between elderly-targeted and
youth-targeted tax expenditures is not nearly so great as in Australia. In
Britain there are signi¬cant tax expenditures on items of interest to the
non-elderly, particularly in the area of labor market supports. A similar
pattern may be observed in Canada, where, as O™Higgins (1988) observes,
the rather average social policy emphasis on young people in the sphere
of direct expenditures is countered by generous tax policies in the area of
labor market supports and family allowances.
Scholarship on tax expenditures in the United States (Longman 1987;
Howard 1997) tends to con¬rm O™Higgins™s assertion that tax expenditures
do not tell a signi¬cantly different story from direct expenditures. The
introduction and expansion of the Earned Income Tax Credit has shifted
the weight of tax policy in the United States somewhat away from the
extreme elderly bias observable in both direct expenditures and the rest
of the tax system. Still, the emphasis on the elderly in U.S. tax policy is
strong, particularly in the ¬elds of housing and private pensions, and cer-
tainly does not counterbalance the extreme elderly orientation of direct
In sum, the best available information on tax expenditures for social pol-
icy points in the same direction as the information on direct expenditures.
With the possible exception of Australia, which also had an ambiguous
ranking based on the ENSR for direct expenditures, tax data con¬rm the
relative age orientation of different welfare states derived from measures of
direct expenditures alone.

Re¬ning the Measure: Housing Policy Patterns and Outcomes
The ¬nal re¬nement of the measure of age orientation refers to the housing
sector. As noted above, tax expenditures on housing and housing-related
debt are, in most OECD countries, the largest tax expenditure on individ-
uals. At the same time, direct public expenditures on housing are relatively
meager. In fact, housing policy in OECD countries is carried out through a
wide variety of policy instruments, ranging from local zoning regulations to
intervention in credit markets to contractor and developer subsidies to land
purchases to direct housing allowances to taxation of imputed rent. Since
many of these policies work in opposite directions, it is dif¬cult to develop

Measuring the Age of Welfare

a measure of the age orientation of housing policy based on regulations and
statutes alone.
But housing policy is an important component of social welfare policy,
both because of its direct effects on quality of life, and because of its impli-
cations for lifetime savings and attitudes toward other welfare programs
(Kemeny 1980, 1981; Castles and Ferrera 1996). Most comparative welfare
state researchers abandon the search for a comparative measure of hous-
ing policy, instead using a single quantitative measure of housing policy
outcomes: aggregate levels of home ownership. This measure is problem-
atic, however, for two reasons. First, and most obviously, aggregate home
ownership statistics obscure differences in home ownership rates among
different age groups in the population. Second, an emphasis on ownership
rates alone ignores the extent to which home ownership is promoted as the
most desired form of housing tenure.
Using housing tenure data from the Luxembourg Income Study (LIS),
I evaluate how well governments live up to their stated housing goals, and
how this varies across age groups in the population. I derive the country™s
housing policy goals from secondary literature (Boleat 1985; Balchin 1996)
and from the responses of housing policy of¬cials to a survey conducted by
the European Union (EU) on housing policy priorities (Dumon 1992). For
some countries, the housing policy priority is to promote home ownership
among the widest possible swath of the population. For others, the priority is
to guarantee a minimum of fairness in the rental sector, either through direct
public provision of rental housing or through protection of renters™ rights
in private markets. Table 2.11 shows the policy focus (home ownership vs.
rental), overall home ownership rates (including cooperative housing), and
the difference in home ownership rates among elderly (over 55-year-old)
and young (25- to 34-year-old) adults, for those countries for which data
were available.
What do these home ownership outcomes imply about housing policy
inputs? I assume that home ownership rates among different age groups
are determined by a range of housing policy inputs, including government
regulation of credit markets and policies that increase the availability of low-
cost homes for private ownership, increase the availability of low-interest
and low-down-payment loans for ¬rst-time home buyers, and encourage
home ownership through ¬scal instruments targeted at lower income home-
owners. These kinds of policies will increase levels of home ownership
among young people, who tend to be asset-poor and income-poor relative

Age in the Welfare State

Table 2.11 Housing policy orientations (late 1980s to early 1990s)

Aggregate Elderly-youth Aggregate Elderly-youth
home difference in home difference in
Policy ownership ownership Policy ownership ownership
focus rate rates focus rate rates
NET Rent 33.7 3.60 LUX Own 69.6 21.5
DEN Rent 59.9 5.50 BEL Own 68.8 21.8
SWE Rent 57.7 6.10 FIN Own 75.2 23.5
GER Rent 42.8 15.6 CAN Own 68.0 27.4
AUT Rent 49.7 16.3 FRA Own 59.1 33.7
AUS Own 73.9 33.9
US Own 66.7 37.7
SPA Own 72.2 41.1
ITA Own 59.1 41.8
Sources: Policy focus and aggregate home ownership: Balchin 1996; Dumon 1992; elderly-youth
difference: author™s calculations from LIS data (Wave IV).

to older people. These types of policies will thus tend to reduce the dif-
ferences in relative levels of home ownership between the young and
the old.
Table 2.11 reveals that countries with similar housing policy goals vary
substantially in the degree to which these goals are achieved for different
age groups. For example, among those countries where home ownership
is not a stated priority, Austria and Germany stand out for the large differ-
ences in home ownership rates between younger and older populations. The
Netherlands, Denmark, and Sweden, on the other hand, as in other areas of
social policy, show more balanced results for different age groups. Among
countries where home ownership is an explicit goal of housing policy, the
United States, Spain, and Italy clearly have achieved that goal to a much
greater extent for their elderly citizens than for young people. And although
I was not able to calculate home ownership rates by age group for Japan,
Boleat (1985, 403) reports a similar age variation in tenure: overall, 60 per-
cent of Japanese households own their homes, while this is true for only
17 percent of households headed by persons under 29, and 46 percent of 30-
to 49-year-olds. Belgium, Finland, and Canada show differences in owner-
ship rates that probably re¬‚ect these countries™ efforts to encourage home
ownership among young people. Again, home ownership patterns support
the picture painted by the basic ENSR: the United States, Austria, Japan,

Measuring the Age of Welfare

Italy, and Spain tend to have among the most elderly-oriented housing pol-
icy regimes, while the Netherlands, Sweden, and Denmark have among the
most age-neutral housing policies. Once again, Australia ¬ts uneasily into
the overall scheme, and we lack data for the United Kingdom, Ireland, and
New Zealand that might help to con¬rm whether the age orientation of
housing policy, as in direct expenditures, is a dimension that cuts across the
traditional Liberal welfare state group.

This chapter has presented a variety of measures of the age orientation
of social policy, based on direct expenditures, tax expenditures, and hous-
ing policy. While each measure presents a slightly different picture, taken
together they reinforce one another. This triangulation of measures permits
us to conclude with some con¬dence that countries do vary in the amount
of emphasis they place on helping their elderly versus non-elderly popu-
lations through public social policies, and vary in consistent ways. In par-
ticular, we note that the most youth-oriented social policy regimes belong
to the Scandinavian and British Commonwealth countries, while the most
elderly-oriented countries are a more diverse group, encompassing parts
of Continental Europe (Italy, Greece, Spain, Austria), the United States,
and Japan. Furthermore, the ENSR measure based on direct expenditures
alone appears to approximate rather well the age orientation of social policy
more generally across countries.
O™Higgins (1988) identi¬es a generalized pattern in OECD countries of
expansion of welfare bene¬ts for families in the 1950s, with retrenchment
in these areas and growth in the pension sector from the 1960s through the
mid-1980s. His ¬ndings accord with Thomson™s (1993) hypothesis that a
“sel¬sh generation” has captured welfare policy across the OECD, design-
ing welfare states to meet the needs of their steadily aging cohort. However,
the data presented here show much greater variety in social policy orien-
tation than is suggested by O™Higgins or Thomson. While the elderly bias
in some countries is indeed acute, in other countries younger age groups
enjoy signi¬cant bene¬ts “ although whether this has occurred through the
political action of age-based constituencies or as a result of other processes
remains to be seen.
The remainder of this book focuses on identifying the causal processes
that generate the diversity of public policy orientations toward different age

Age in the Welfare State

groups that we have observed here. To what extent are differences in the
age orientation of social policies the result of conscious policies designed
to privilege certain age groups or generations over others? How much do
these policy differences re¬‚ect societal attitudes about the relative neediness
or deservingness of different age groups? Do they in fact spring from the
interaction of political actors seeking to protect interests that are de¬ned
by age at all?


Age and the Welfare State

Chapter 2 revealed wide variation among industrialized countries in the rel-
ative emphases that governments place on social protection for elderly and
non-elderly population groups. But the age orientation of social policies
varies across countries in ways that are quite unexpected given what the
scholarly literature on comparative social policy tells us about how welfare
states develop. Esping-Andersen™s (1990) three welfare regime types do
not separate neatly into youth-oriented, elderly-oriented, and age-neutral
welfare states, as we might expect. And basic country attributes such as
aggregate levels of welfare spending or the size of the elderly population
tell us even less about the probable age orientation of a given welfare state.
This chapter looks systematically at a variety of potential explanations for
why the industrialized countries in this study display such different social
policy age orientations. Scholars have called on a variety of structural, cul-
tural, political, and institutional features of nation-states to explain diver-
gent patterns of welfare state development. These existing theories about
welfare state development were not designed explicitly to explain the age
orientation of social policies. But, as we have seen, age orientation is a
fundamental aspect of how welfare states redistribute resources, with con-
sequences for labor and ¬nancial markets, family structures, fertility, and
so on. So it is fair to expect that these classic explanatory paradigms should
be able to account for this important aspect of what welfare states do and
how they do it.
Yet existing paradigms do not perform well when confronted with the
task of explaining variation in the age orientation of welfare states. And
with good reason. The age structure of the population, ideologies about
redistribution across the life course, and the political power of groups with
age-related policy agendas fail to explain variation in the age orientation of
Age in the Welfare State

welfare states precisely because the age orientation of social policies is not,
in fact, related to social structural, partisan, or institutional features that are
in any straightforward way linked to age. Rather, the distinct age pro¬les
of social policy regimes are a largely unintended consequence of how wel-
fare state programs are structured, and how politicians typically compete
within a party system. The second half of the chapter presents an alternative
explanation for the variation in age pro¬les of social policy regimes, arguing
that early choices about the structure of welfare programs combine with
distinctive modes of political competition in different countries to account
for the development over time of differing age orientations.

Classical Explanations for Variation in Welfare State Outcomes
The comparative political economy literature has elaborated a variety of
explanations for why welfare states vary in their form and function, some
of which could yield insights into the sources of cross-national variation in
the age orientation of social policies. Modernization approaches identify eco-
nomic and demographic “development” as the keys to understanding how
much, and what, welfare states do. One subtype of this argument deserves
special attention: the political impact of pensioner lobbies and elderly vot-
ers, “gray power,” has been hypothesized to affect welfare state spending in
important ways. Welfare state familialism, often linked to the preferences of
Christian Democratic political actors, may also be related to the age orien-
tation of social policies. Power resources, namely, the strength of class-based
actors such as Social Democratic parties, unions, and employers, tell us a
great deal about why welfare states vary along dimensions not related to
age. To the extent that social spending on different age groups has equity
implications, left power resources may also be important predictors of the
age orientation of welfare states. Employer preferences, the subject of a
current wave of scholarship in comparative political economy, may also
explain why welfare states develop a particularly elderly-oriented or youth-
oriented repertoire of social policies. Finally, institutionalist approaches that
argue for the impact of constitutional structures, forms of interest interme-
diation, or state capacities all offer insights that could help explain why the
age orientation of social policies varies from country to country.
Each of these classical approaches can add something to our understand-
ing of why some countries devote the lion™s share of their welfare resources
to the elderly, while others focus more on the needs of working-age adults
and children. A fuller explanation of the age orientation of welfare states,
Theories and Hypotheses

though, requires attention to both the political and institutional contexts
within which political entrepreneurs forge links to potential constituencies
of the welfare state. The second part of this chapter lays out such an expla-
nation. First, let us consider how far classical welfare state theories can take
us toward understanding why nations vary in the age orientation of their
social spending.

Developmentalist approaches to explaining the growth of the welfare state
(see, e.g., Wilensky 1975; Flora and Heidenheimer 1981; Flora and Alber
1983; Flora 1986; Myles 1989) argue that the arrival of industrial soci-
ety creates both new social needs and the resources with which to meet
these needs. Of special relevance to the problem of understanding the age
orientation of welfare states is the contention in many such accounts that
industrial capitalism creates particularly pressing needs among the elderly,
as it creates a new class of inactive elderly persons: retirees. This devel-
opment in turn drives the expansion of bene¬ts for the elderly. Accord-
ing to this logic, welfare states develop primarily to protect the elderly
because the needs of the elderly are (1) the ¬rst to emerge on a large
scale in industrialized society and (2) the most remote from the tradi-
tional concerns of care-giving institutions such as the family, poor laws,
or private charity. Such assumptions are in fact echoed in policy discus-
sions surrounding the implementation of pension schemes, where debate
often focuses on the uniquely deserving character of the elderly or on
the impracticality of relying on families to provide the necessary income
Welfare state policies do not, however, appear in practice to develop in
response to a temporal primacy of the needs of elderly people. It is true that
most industrialized countries adopted old-age insurance programs before
either unemployment insurance or family allowances. But in most states the
¬rst public welfare bene¬ts to be introduced were either poverty alleviation
programs, targeted at children and adults as well as the elderly, or sickness
and occupational injury insurance bene¬ts for current workers (Flora and
Alber 1981). Neither do bene¬ts for the elderly come ¬rst in the sense of
being a higher priority. If this were the case, we would expect countries
with lower levels of aggregate welfare spending to be the most elderly-
oriented, while only high-spending welfare states would be able to afford
the “luxury” of youth-oriented social programs. However, elderly-oriented
Age in the Welfare State

countries with high levels of social spending as a percentage of GDP (e.g.,
Italy) coexist with small, youth-oriented welfare states (e.g., Ireland).

Gray Power
Even if there is no mechanical correspondence between economic and
demographic development and elderly-oriented welfare spending, it still
seems plausible that the age structure of populations could affect the direc-
tion of welfare state spending via the political power of the elderly. Some
theorists argue that growing state spending on pensions is a result of the
in¬‚uence of gray power: large blocs of elderly voters with well-de¬ned pol-
icy preferences. In one of the ¬rst quantitative cross-national studies of the
welfare state, Wilensky (1975) argued that elderly populations in¬‚uence
the development of welfare state spending because large elderly popula-
tions create both a need for more welfare spending and a political con-
stituency to ¬ght for the allocation of resources. Pampel and Williamson
(1989) likewise found that in democratic countries the “political pres-
sure of a large aged population” is an important in¬‚uence on spending.
Thomson (1989) posited the aging of a politically powerful “welfare gen-
eration” as the driving force behind the growing emphasis of welfare states
on programs for the elderly versus programs for children from the 1970s
More recently, Pierson (1994) and Campbell (2003) examine the impact
of the elderly on welfare state retrenchment. Both argue that policy legacies
shape the interests and capacities of elderly constituencies of the welfare
state, and in so doing constrain subsequent policy making. Pierson speci¬es
a set of micro-foundations underlying his argument about policy legacies.
Voters exhibit a powerful negativity bias, ¬ghting retrenchment of “their”
programs at almost any cost; and politicians use strategies of blame avoid-
ance to get around the resistance of constituencies to policy retrenchment.
These micro-foundations suggest that the preferences of numerous and
highly motivated elderly voters regarding pension programs lie at the heart
of the politics of welfare reform.
But gray power still falls far short of predicting when, where, and why one
might observe a bias toward the elderly or the young in welfare spending. If
the size of the elderly population, and thus its electoral strength, were a good
predictor of the age orientation of social spending, we would expect to see
all countries becoming more elderly-oriented with the passage of time. But
analysis of spending data over time reveals that nine of the twenty countries
Theories and Hypotheses

Table 3.1 Percent change in ENSR, 1960“2000

Country Percent change since 1960
New Zealand
Greece 17
France 23
Italy 42
Sweden 49
Spain 50
Norway 76
Belgium 86
United Kingdom 92
United States 104
Canada 116
Japan 799
Notes: The reference period for Portugal is 1977“2000; Greece
1962“2000; Spain 1967“2000; Belgium and U.S. 1960“99; U.K.
1960“98; and Japan 1970“2000.
Sources: Spending data: OECD 2004; demographic data:
OECD 2003b.

in this study have in fact become more youth-oriented between 1980 and
2000 (see Table 3.1).
Why does the gray power hypothesis fall so far short of the expecta-
tions generated in the literature on comparative welfare state spending?
The ¬rst reason is an almost trivial one: there is no general agreement
on what kind of welfare spending the elderly should, let alone do, prefer.
Wilensky (1975) and Pampel and Williamson (1989) originally surmised
that the elderly would favor elderly-oriented spending. But in 1990 Wilen-
sky reversed himself, asserting that elderly voters are in fact more altruistic
in their policy preferences and are naturally inclined to support state sub-
sidies for families with children (Wilensky 1990). By 1993 Williamson and
Pampel were likewise less convinced that the elderly would support more
pension spending in all political contexts. The disagreement about what
Age in the Welfare State

the elderly want could be resolved empirically, but the existing literature
on this subject is rather thin outside the United States.
A second problem with the gray power approach is less trivial. It assumes
that voters know what they want and communicate their desires upward to
politicians, who then act on their constituencies™ policy preferences. Pier-
son™s (1994) work, for example, casts doubt on this model, but ultimately
does not reject it. Pierson has a persuasive account of how policy feedback
and politicians™ behavior affect voters™ behavior. He emphasizes both the
demobilizing effects of politicians™ efforts to frame policy changes in a pos-
itive light and the capacity of earlier policy decisions to shape how welfare
state bene¬ciaries organize themselves politically. But Pierson remains tied
to a view of political preference formation in which, at any given time, the
bene¬ciary/voter takes stock of a full range of policy options and expresses a
preference, and it is once again up to the politicians to dissuade, dissemble,
and deceive the voters into accepting retrenchment.
A well-established literature exists, however, indicating that policy pref-
erences get their start among political leaders, and are only subsequently
taken up by mass publics. Converse (1964) and Zaller (1992) make the
strongest case for elite leadership of public opinion, but others (Arnold
1990; Gerber and Jackson 1993), too, argue that elites form opinions prior
to their constituencies on at least some issues some of the time. Research
on framing (e.g., Gilens 1999) and agenda-setting (e.g., Baumgartner and
Jones 1993) further suggests that voters do not select their policy prefer-
ences from among a complete set of all plausible options, but rather take
their cues about what is possible and desirable from politicians. The gray
power approach forgets to ask what politicians want, how they communicate
these preferences to the electorate, and how the electorate™s choices are
constrained by the shape of elite opinion and strategy.

Modernization and gray power approaches focus on the politics of need:
welfare states respond to new needs, especially among the elderly, in pro-
portion to the level of need and/or the political power of those who are
(or will be) in need. A different strand of research shifts the focus to the
ability and propensity of other societal actors to provide for needs with-
out recourse to public social programs. Recent scholarship suggests that
deeply held societal values that are re¬‚ected in family structures and reli-
gious orientations can affect welfare state outcomes. Both family structures
Theories and Hypotheses

and religious doctrines could plausibly be held to shape the political posi-
tions of various actors when it comes to issues of intergenerational equity
and the proper role of family versus state in caring for the needy at differ-
ent stages in the life course. In particular, the dominance of Catholic social
doctrine in some Continental European countries and the persistence of
multigenerational family structures in Southern Europe and Japan have
been set forth as explanations for the elderly-orientation of these welfare
Van Kersbergen (1995) argues that Christian social doctrine has distinc-
tive effects on social policy outputs. On its face, the strength of Christian
or Catholic social values in different countries is a weak explanation for
variation in age orientation: Christian democratic countries such as Italy,
the Netherlands, Belgium, Germany, Austria, and Portugal share a cause
(Christian social doctrine) but vary widely on the effect (age orientation).
Still, key tenets of Christian social doctrine, including a focus on fami-
lies as the primary providers of social assistance, and on a family structure
revolving around a male breadwinner, have been mustered to explain both
very high child allowances and unemployment bene¬ts in the Netherlands
(Bussemaker 1992) and very low bene¬ts for young people in Italy (Sara-
ceno 1994). Given these con¬‚icting claims, it seems wise to investigate the
sources from which cultural ideas about appropriate care for different age
groups are drawn. But even a close reading of the Catholic encyclicals most
explicitly dedicated to the social policy issues, Rerum novarum (Leo XIII
1891) and Quadragesimo anno (Pius XI 1931), reveals no preference, explicit
or implicit, for social provision for one age group over another. The empir-
ical evidence that Christian or Catholic cultural values cause distinctive age
orientations in social policies then seems rather thin.
The ideas contained in Christian social doctrine very likely do matter
for welfare state outcomes, and even for the age orientation of welfare
states. They may play an important role in informing the policy choices
pursued by various political actors, including but not limited to Christian
Democratic parties and Church lobbyists. However, the degree to which
Catholic-dominated welfare states differ in their treatment of different age
groups is striking. Thus, the in¬‚uence of Catholic familialism on welfare
state outcomes must be conceived of in terms of the variety of legitimat-
ing ideas about social policy that could plausibly be supported by Catholic
doctrines, and the selection of ideas that eventually become in¬‚uential in a
given polity. In other words, if we wish to understand why Catholic coun-
tries have such different pro¬les of distribution across age groups, we must
Age in the Welfare State

try to ¬gure out why different tenets of social Catholicism are emphasized
in different settings, and what are the pathways by which these values ¬lter
into the policy-making arena.
Other scholarship emphasizes the importance of family structures them-
selves for social policy outcomes (see Jurado and Naldini 1996). According
to many observers of Southern European politics, in particular, the preva-
lence of multigenerational families and a pervasive familialist orientation
in these countries accounts for the underdevelopment of public policies
ranging from child care to social services to unemployment bene¬ts. This
argument could apply to the Japanese case as well, where multigenerational
families are also relatively common.
It is certainly plausible that cohesive extended families engaging in exten-
sive intrafamilial resource sharing make it possible for Southern European
countries to sustain high levels of unemployment without falling prey to
debilitating social con¬‚ict between labor market “insiders” (primarily older,
male workers and pensioners) and “outsiders” (younger, female workers
and the unemployed). It is far from clear, however, that family structures
are the cause of limited bene¬ts for working-age adults and children in
Southern European countries, rather than the other way around. True,
Southern European social policies rely on extended family structures to a
greater extent than in other countries (Millar and Warman 1996; Naldini
2003). But, at least in the Italian case, this is a rather recent phenomenon.
The tendency for social legislation to focus on the family as primary care-
giver and source of income support is, according to Addis (1998) and
Saraceno (1999 interview), a result of increasing demands and decreasing
welfare state resources, rather than a result of the impact of a familialist
Indeed, research in other areas of welfare state provision suggests exer-
cising caution before drawing a direct causal link between extended family
structures, on the one hand, and public policy outputs, on the other. Jurado
(2002) demonstrates that the family structure most characteristic of South-
ern European societies, the long permanence of adult children in their
parents™ households, is caused by characteristics of housing and labor mar-
kets in Southern Europe, rather than by the socio-cultural features more
commonly assumed to be the culprit. Similarly, research on public atti-
tudes toward family policies in Italy reveals that respondents who agree
most whole-heartedly with “traditional” family values are also those who
are most supportive of strong state intervention on behalf of family val-
ues (Palomba 1995). Extended family structures and a Catholic culture
Theories and Hypotheses

emphasizing the subsidiarity principle “ leaving the family to its own
devices “ are not enough to explain the relative paucity of state welfare
provisions for younger people in Southern Europe.

Power Resources
Since the 1980s many of the most in¬‚uential studies seeking to explain vari-
ations in welfare states™ timing, size, structure, and performance have been
grounded in the study of the political power of the working class and, more
recently, employers. Power resources approaches argue that welfare state
outcomes can be explained by the political strength of class-based political
actors. In its more traditional formulation, the emphasis is on working-
class actors alone (see Stephens 1979; Korpi 1983; Esping-Andersen 1985;
Myles 1989). Highly developed welfare states that reduce income inequal-
ity are, in this account, the result of the political strength of representatives
of the wage-earning classes, who use the political arena to combat market
dynamics. More recent work in the power resources tradition highlights the
importance of employers™ preferences (Swenson 2002; Mares 2003) or of
coalitions between working and middle classes (see Baldwin 1990; Esping-
Andersen 1990), but shares with earlier analyses a focus on how the power
of class-based actors affects policy outcomes.
Whether left power resources can adequately explain the observed varia-
tion in social policy age orientations is ultimately an empirical question. All
other things being equal, we might expect working-class political actors not
to prefer elderly-oriented social spending, which concentrates bene¬ts on
one group rather than spreading them in a more egalitarian fashion across
the population. But in practice, the Left has often preferred programs that
are highly decommodifying for older citizens but not younger ones, espe-
cially when the alternative is a welfare state that is equally mean toward all.
Social programs linked to occupational performance tend to generate more
elderly-oriented spending than citizenship-based programs do, and Social
Democratic and union actors in some contexts now prefer universalist social
policies to occupationalist ones. But there is no reason to think that politi-
cal actors on the Left had this in mind when they advocated for particular
program designs. In fact, there is ample evidence that the very nonoccu-
pational welfare state structures that have over the course of the twen-
tieth century matured into rather youth-oriented systems were imposed
against the will of Social Democratic actors (see Baldwin 1990; Swenson
Age in the Welfare State

We can ask a similar set of questions about employers. What kinds of
policies do they prefer? What would these policies imply for the age ori-
entation of social spending? And did employers in fact get the policies they
wanted? One hypothesis suggests that relatively youth-oriented policies in
the small, open economies of Scandinavia, the Benelux countries, Austria,
and Ireland are the result of employers™ distinctive preferences for univer-
salist social policies in these countries. While increasing economic openness
within countries over time does not uniformly coincide with more youth-
oriented social spending (see Fig. 3.1), it is worth exploring the possibility
that there might be a causal connection between the preferences of employ-
ers in small, open economies and the age orientation of the social policies
in these countries.
Employers™ product-market strategies (Estevez-Abe, Iverson, and
Soskice 2001) are collinear neither with economic openness nor with age
orientation. But it seems plausible that employers in small, open economies
might prefer universalistic, tax-¬nanced social programs to occupationally
based programs. The former would reduce employers™ direct nonwage labor
costs, while the latter would impose costs that employers could not pass on
to consumers without decreasing their international competitiveness. Let
us for a moment accept the premise that, on balance, employers in small,
open economies will prefer citizenship-based social policies to occupational
social insurance. Demonstrating that employers™ preferences are responsi-
ble for the age orientation of welfare states in these countries then relies
on showing that employers™ preferences were decisive in both the decision
to introduce citizenship-based social policies in the small, open economies
around the turn of the twentieth century and the decision to introduce
new universalistic programs in those countries that had occupational social
insurance systems entering World War II.
The ¬rst of these two claims is plausible, at least for the Scandinavian
countries. Export-oriented (agrarian) employers in Denmark, Sweden, and
Finland seem to have preferred citizenship-based programs. And in alliance
with Liberals, their policies won out over those of the Social Democrats,
who advocated occupationally based bene¬ts (Baldwin 1990, chapter 1;
Kangas, in press). The second claim, that a shift to more universalistic
social policies in some countries after the Second World War resulted from
employers™ preferences in these countries, is less tenable.
As Katzenstein remarks in his authoritative tract on the political
economies of small states, “domestic compensation . . . responds pri-
marily to the logic of domestic politics; it is not a deliberate response to






Percent change, 1960-2000




change in openness change in ENSR

Figure 3.1 Change in economic openness and age orientation, 1960“2000. Note: Openness is ratio of
exports plus imports to GDP. Germany openness is for 1970“2000; Greece ENSR is for 1962“2000; Japan

ENSR is for 1970“2000; Portugal ENSR is for 1977“2000; Spain ENSR is for 1967“2000; UK ENSR is
for 1960“98; and U.S. and Belgium ENSR is for 1960“99. Sources: Openness: Heston, Summers, and Aten
2002; ENSR: spending data: OECD 2004; demographic data: OECD 2003b.
Age in the Welfare State

the logic of the international economy” (1985, 133“4). A shared open-
ness to international markets certainly cannot explain why Austrian and
Belgian employers might have preferred to compensate for economic open-
ness via occupationalist welfare programs, while Scandinavian employers
chose universalist ones earlier in the century. Furthermore, if Katzenstein
is correct that employers in Belgium and Austria were more able than in
Scandinavia to impose their demands on “a labor movement too weak
to dictate its own terms” (173), it then follows that occupationalist pro-
grams in Belgium and Austria are more plausibly the result of employers™
The genesis of the initial split between occupationalist and citizenship-
based welfare states in the early twentieth century is less at issue here than
the persistence of these institutional choices over the course of the next cen-
tury. The claim that employers might have been responsible for the shape
of social policies in small, open economies at the turn of the century tells
us little about this persistence, which, as we shall see, is crucial to the even-
tual age orientation of welfare states. It is possible that in some countries
where citizenship-based social policies are of long standing (e.g., Sweden),
employer interests, reaf¬rmed over the course of the twentieth century,
locked these policies into place (for such an argument, see Swenson 2002).
But the case studies of social policy development in the Netherlands in the
post“World War II period (chapters 4 through 6) do not support the inter-
pretation that employers™ interests drove the transition from occupationalist
to universalist policies in that particular small, open economy. Generalizing
from either Sweden or the Netherlands to the effects of employers™ pref-
erences in all of the small, open economies seems risky in the absence of
detailed cross-case historiographical evidence.
Regardless of whether the focus is on the working class or on employ-
ers, the very logic of power resources analysis brings to the fore issues that
are helpful in formulating alternative hypotheses. The power resources
approach assumes that the policy preferences of class-based actors are
deducible from their positions in the productive system. But the interests of
wage earners and employers can be powerfully affected by a variety of other
conditions. Mares (2001), for example, argues that employers™ preferences
about pension reform are conditional on the percentage of social expendi-
tures currently allocated to pensions. It is reasonable to presume that savvy
class representatives, like all good politicians, will advance policy demands
that seem feasible and likely to produce desired results given the extant polit-
ical and policy environment. Class interests in the abstract thus seem to be a
Theories and Hypotheses

poor predictor of the policy demands that are likely to emanate even from
relatively homogenous class-based political actors.
Power resources approaches also tend to assume an unrealistic degree of
homogeneity of interests within working-class-based organizations. This
is especially relevant for the study of age-related spending priorities, since
most working-class organizations encompass both older and younger work-
ers. Parties and unions are cross-age coalitions, and as such may adopt
contradictory or difference-minimizing positions on issues related to inter-
generational distribution in an attempt to hold together overlapping class-
and age-generated cleavages (see Anderson and Lynch 2003; Natali and
Rhodes 2004). So even if we could deduce the age-related policy prefer-
ences of working-class organizations from their class origins (which seems
unlikely), the internal dynamics and external environments of parties and
unions would still affect the welfare policy positions that they advocate.

Political institutions constitute one important aspect of the environment
for class-based (and other) political actors. A range of political institutions,
from neo-corporatist bargaining structures to electoral systems to judicial
review, have been hypothesized to affect the development of welfare poli-
cies. Within the comparative social policy literature, attention has been
focused on two main types of institutions: constitutional structures (so-
called veto points) and neo-corporatist industrial relations.
The literature on constitutional structures (see, e.g., Imergut 1992;
Huber, Ragin, and Stephens 1993) argues that features of a nation™s for-
mal institutional landscape can determine social policy outcomes by setting
up rules of the game that favor certain political actors over others. The
expected consequences of different constitutional structures for the age
orientation of social policy regimes are not immediately clear. Still, there
seems to be no a priori reason to reject the hypothesis that these sorts of
institutions could matter. But such a static vision of institutions offers little
hope for understanding why the age orientation of social policies changes
over time within countries.
There is empirical support for the idea that neo-corporatist institutions
may affect the age orientation of social spending (Pampel 1994). But the
literature on corporatism and social policy makes bifurcated predictions
about the consequences of incorporating organized interest groups directly
into policy making. On the one hand, optimists (see, e.g., Katzenstein 1985;
Age in the Welfare State

Visser and Hemerijck 1997) view positively the capacity of corporatism to
result in policies that are in the general interest, and thus not particularly
oriented toward one age group or another. These authors predict that in the
presence of corporatist institutions, social policies will be other-regarding,
promoting equity across wide segments of the population and compensating
societal losers. These bene¬cial results occur, according to these authors,
because neo-corporatist policy-making processes enhance possibilities for
trust, long-term engagement, and positive-sum games. Other analysts (e.g.,
Offe 1981; Olson 1982; Esping-Andersen 1996) express more pessimism
about the policy outputs of corporatism. For these authors, corporatism can
perniciously link the inherently rent-seeking behavior of organized inter-
ests to policy making. This results in public policies that protect labor™s,
employers™, or welfare constituencies™ interests narrowly de¬ned, but that
do little to advance equity, competitiveness, or long-term economic perfor-
mance. This viewpoint suggests that corporatism may enhance the capacity
of powerful elderly groups to pursue their own policy agendas on aging, to
the detriment of other age groups in the population.
Quantitative studies of welfare state outcomes, including aggregate
spending, spending in particular program areas, and income inequality,
generally support the notion that corporatist institutions result in bigger,
more egalitarian welfare states (see, e.g., Esping-Andersen 1990; Hicks and
Swank 1992; Hicks and Misra 1993; Birch¬eld and Crepaz 1998; Crepaz
1998; Bradley, Huber, Moller, et al. 2001). Pampel (1994) even puts forth
evidence suggesting that corporatism has a distinctive youth-oriented effect
on welfare spending. But the measures of corporatism employed in quan-
titative cross-national studies leave a great deal to be desired. Scholars who
have attempted to de¬ne and measure corporatism cross-nationally (see,
e.g., Schmitter 1981; Wilensky 1981; Cameron 1984; Lehmbruch 1984;
Crouch 1985) disagree about the core concepts that should be included in
the term and their range of applicability, resulting in important differences
in how different countries are scored from one measure to the next. And
many welfare state scholars use measures that do not capture changes over
time in the degree or kind of corporatism present in a particular coun-
try. Detailed process-tracing analyses of the kind presented in chapters 4
through 6 of this volume are necessary to understand the link between
corporatist bargaining structures and the age orientation of social policies.
The Dutch and Italian case studies hint at some connections. In Italy,
episodes of tripartite concertation seem to strengthen the hand of broad-
based forces within the union movement that have little to lose and much
Theories and Hypotheses

to gain from reining in pension spending. This mechanism may be gen-
eralizable to other contexts (Anderson and Lynch 2003 propose such a
model). However, more careful qualititative research is necessary to con-
¬rm whether either Pampel™s (1994) model or Anderson and Lynch™s ¬ts
the historical evolution of spending in the countries with universalist social
The literatures on veto points and corporatism hold that formal insti-
tutions “ in some cases constitutional provisions or laws, in other cases
government-sponsored agreements rati¬ed by trade union and employer
organizations “ affect the age orientation of welfare provisions by enhancing
or reducing the bargaining power of particular political actors. A different
strain of institutionalism focuses on less formal institutions and on the ways
in which these institutions affect both the range of possible policies and
the preferences of different political actors. An exemplary work in this vein
is Orloff ™s The Politics of Pensions (1993). Orloff argues that three kinds of
institutions are particularly important in shaping the development of pen-
sion policies in the United States, the United Kingdom, and Canada: state
capacities, in particular the ability to tax; the mode of operation of politi-
cal parties and bureaucracies, either patronage-oriented or programmatic;
and feedback effects from earlier policy decisions. Orloff ™s argument is
compelling and anticipates some of the most interesting insights regarding
policy feedbacks from Pierson™s (1994) work on welfare retrenchment in
the United States and the United Kingdom. But it is highly contextualized
and dif¬cult to generalize. The core argument of this book concurs with
Orloff ™s in important respects, including its focus on ¬scal capacity and the
modus operandi of political parties. However, it generalizes these results
beyond pension policy, beyond the Anglo-Saxon countries, and beyond the
early twentieth century, allowing us to see the results of particular institu-
tional con¬gurations for social policy outcomes in a wide variety of national
settings. This book also takes up the gauntlet thrown down by Thelen
(1999) and Pierson (2000), investigating closely how institutions repro-
duce themselves over time and, in so doing, create enduring social policy

A Path-Dependent Institutionalist Explanation
The remainder of this book argues that path-dependent political and social-
policy institutions are the best explanation for the age orientation of welfare
state spending. How social programs are organized (along citizenship or
Age in the Welfare State

occupational lines) and how politicians compete with each other (program-
matically or using particularistic appeals to groups of voters) are the key
factors that determine patterns of social spending on different age groups.
But how are these two factors related to one another, and ultimately to the
age orientation of social spending?
To preview: at two critical junctures the welfare states of the industrial-
ized democracies set out on trajectories toward divergent age orientations.
From the ¬rst critical juncture, in the early twentieth century, two groups
of countries emerge with welfare states that are organized according to
radically different logics, either citizenship-based or occupationally based.
These organizational forms mature into welfare states of different hues.
Citizenship-based programs become more youth-oriented with the pas-
sage of time, while occupationalist programs contain within them the seeds
of elderly-oriented social spending. But in order for these divergent age ori-
entations to develop, countries must maintain their institutional setups well
into the postwar period. At a second critical juncture, around the Second
World War, the occupationalist camp divides into two further groups: one
that maintains occupationally based family allowance and unemployment
programs and one that replaces many of its prewar occupationalist programs
with citizenship-based ones. This second parting of ways is explained and
reinforced by the predominant mode of political competition in these coun-
tries, either programmatic or particularistic. The outlines of this argument
are illustrated in the form of a branching tree in Figure 3.2.
The structure of core welfare state programs is clearly correlated with
the age orientation of welfare states, for reasons that shall become clear
shortly. Countries that have universal, citizenship-based provisions for old
age, unemployment, and child rearing tend to be more youth-oriented.
Quite surprisingly, this is true regardless of the overall size of the welfare
state relative to GDP, regardless of whether programs are means-tested
or not, and regardless of whether the basic citizenship-based bene¬t is
supplemented by a public occupationalist tier. Occupationally based social
programs, on the other hand, tend to generate elderly-oriented welfare
states. But the structure of welfare programs is only a partial explanation
for why some countries treat some age groups more generously than others.
Understanding how the observed correlation between program structure
and age orientation develops over time requires identifying the “repro-
duction mechanisms” (Thelen 1999) that reinforce the choice of program
structure entered into at a particular critical juncture. Without this knowl-
edge, we cannot understand why and how early choices about welfare state
Theories and Hypotheses

First Great Divide (ca. 1900)

Citizenship-Based Occupational
Welfare Regimes Welfare Regimes

Second Great Divide (ca. WWII)

Programmatic Particularistic
Competition Competition

Universal and Mixed Occupational
Means-tested Systems Systems Systems


Least Elderly-Oriented Most Elderly-Oriented
Figure 3.2 Watersheds of welfare state formation.

institutions play out over the long run to produce the age orientations we
observe today. How political competition is organized in different countries
helps to explain why occupational programs persist in some countries but
not others “ and in turn why some countries with similar welfare program
structures in 1900 ended up with very different age orientations at the end
of the century.
Age in the Welfare State

Citizenship-Based versus Occupational Programs
Let us begin with an assertion: that welfare regimes with occupationalist pro-
grams produce elderly-oriented social spending, while welfare regimes with
citizenship-based programs produce social spending that favors the young.
We can think about welfare state regimes as lying along a continuum
according to the structure of their main social programs. At one end of the
continuum we ¬nd pure citizenship-based regimes. Here, welfare bene¬ts
may be either means-tested or truly universal, but in either case eligibility
for these bene¬ts is the same regardless of the person™s job title, sector, or
labor market status. All citizenship-based systems cover relatively young
labor market “outsiders” such as mothers and children, which accounts in
large part for their relative youth orientation. Universalist citizenship-based
systems also provide protection for workers and pensioners, while means-
tested systems typically leave labor-market insiders to procure insurance in
the market.
On the other end of the spectrum are pure occupational regimes, in which
eligibility for and/or the quality of a full spectrum of social bene¬ts varies
in accordance with a person™s connection to the labor market. Occupational
regimes are elderly-oriented because they focus on providing coverage for
labor market “insiders.” These current or former members of the core work
force constitute an aging subpopulation, because of both the increasing
dif¬culty of absorbing younger workers into well-protected sectors of the
economy and the increasing lifespan of pensioners.
Figure 3.3 shows twenty welfare states divided into three groups based on
the structure of their main pension, family allowance, and basic social assis-
tance bene¬ts in 1970: predominantly citizenship-based, predominantly
occupational, or mixed.1 Countries such as Sweden and the United King-
dom, which top off basic citizenship-based bene¬ts with a much smaller
occupational tier, are classi¬ed as citizenship-based. These countries are
then ranked according to the mean age orientation of their social spend-
ing for the period 1985 to 1998. The relationship between welfare state
structure and age orientation is clear: citizenship-based welfare regimes
are more youth-oriented than occupationalist systems, while welfare states


. 2
( 7)